How Might Artificial Intelligence Help CAD Users?

SOLIDWORKS presentation explains the basics of the AI landscape, and some of the ways that the company is embracing the future of human–machine partnerships.

“I don’t have the time, I don’t have imagination, I am too busy — please tell me what is the best, the optimal shape to achieve my engineering goal.” That’s the kind of request, said SOLIDWORKS CEO Gian Paolo Bassi, that a user might eventually pose to xDesign, the company’s browser-based 3D CAD solution that’s currently in beta. “Armed with that knowledge, it is very easy for me to create [my intended] object,” Bassi continued.

Although the CAD community has been inundated in recent years with talk of artificial intelligence (AI), machine learning, and the like, there’s not always a clear link between the two worlds of technology. What are the various types of AI technologies, and how can they really help CAD users?

According to David Randle, senior business development manager for SOLIDWORKS, “the vision is to empower engineers and designers to solve problems intelligently,” using the definition that “intelligence is comprehension and action working cyclically, yielding progress” — and it can be artificially accelerated.

There’s often confusion about the terminology used in this area. Randle sought to clear things up in his presentation at SOLIDWORKS World 2018, where he described the types of AI — which is the umbrella term that encompasses all these technologies — as follows:

• Machine learning: A system that can learn and improve based on what it experiences, instead of having to be explicitly programmed.
   
• Neutral language processing (NLP): The system has the ability to understand human language as it is spoken; this gives rise to tools that communicate with machines and have them interpret our speech.
   
• Robots and robotic agents: Agents working on their owner’s behalf without interference from that ownership entity. In other words, they don’t require oversight to operate. Randle pointed out that this category could include software, as well as the machines we typically think of as robots.

Existing SOLIDWORKS Integrations

AI is not the stuff of the far-off future; it is already incorporated within the SOLIDWORKS portfolio to support a variety of design goals. Within the category of optimization, AI helps generate more intelligent designs that are lighter, stronger, and more economical, as is the case with Live Parts, which “grows” part models in a manner that replicates living cells. “There’s a lot we can learn from nature,” Randle commented.

In addition to mimicking natural forms and modes of creation, AI can suggest design variants based on goals and constraints, and even automate design. One example is the DriveWorks design automation tool for SOLIDWORKS, which can create 3D models, 2D drawings, and documentation based on data entered into forms. 

In the realm of manufacturing, AI can improve processes by:

• Recommending methods for quicker production times
   
• Suggesting options that will result in cost savings, just as alternate materials or slight changes to features
   
• Determining the best method of production and supplier for each job
   
• Tracking and coordinating predictive maintenance tasks.

A new addition is a tool in SOLIDWORKS Visualize called the AI Denoiser, which has been trained with NVIDIA machine learning technology to anticipate, recognize, and eliminate visual noise during rendering processes. The result is that Visualize can produce finished renderings as much as ten times faster, according to SOLIDWORKS.

The new AI Desnoiser tool for SOLIDWORKS, shown in use in the right halves of these images, speeds creation of renderings by identifying and eliminating noise from the images.

Safety is an element of manufacturing processes not often thought about early in a project, said Randle, but AI can help change that. Based on what would be relevant for the job at hand, AI could help select tests and determine whether all necessary testing was being performed. It could also draw designers’ attention to known points of failure, and recommend design changes based on previous test outcomes.

As Bassi said, “Automation will help you … make better decisions in two ways: by embedding in your flows the power of simulation in real time, everything that you do needs to be simulated with the most realistic outcome. And … through introspecting the knowledge that is now gated in millions of parts, millions of design decisions that you made through the years.”

Another mass of data that’s potentially very valuable, but difficult for humans to sort through, is sensor data of all kinds. “There’s already so much content, and discovering it is very difficult,” Randle said. AI makes it possible to find the sought-after needle in a very big haystack; Randle pointed to Google Photos as one familiar example.

And to calm those who fear the loss of their jobs in the coming age of robotic modelers and drafters, Randle cited a Gartner prediction that the 1.8 million jobs to be displaced by AI will be outweighed by the 2.3 million new ones it will create. “There’s going to be more opportunity with the introduction of these jobs, not less,” he said. “AI cannot replace a human in any part of this process yet … I don’t see that happening soon.” Instead, he said, it will replace aspects of workflows that are not enjoyable, presumably freeing up humans for more fulfilling and interesting tasks.